论文部分内容阅读
利用反演分析得到的参数进行高面板坝的应力、变形分析来预测长期变形。由于堆石坝的施工过程和变形机制比较复杂,很难将瞬时变形和流变变形分开,因此,有必要对静力本构模型参数和流变模型参数进行综合反演。利用实测位移资料,以对堆石坝变形较敏感的静力本构模型和流变模型参数为待反演参数,采用基于粒子迁徙的粒子群算法和径向基函数神经网络构建参数反演平台,该方法克服了粒子群算法易陷入局部最优和早熟收敛的缺点,采用经过训练的神经网络来描述模型参数和位移之间的映射关系,节省了参数反演的计算时间。对水布垭高面板坝的反演结果表明,基于反演参数的沉降计算值与实测值吻合得很好,坝体变形在合理范围以内并趋于稳定。
Using the parameters obtained from inversion analysis, the stress and deformation of high dam are predicted to predict the long-term deformation. Due to the complicated construction process and deformation mechanism of rockfill dam, it is very difficult to separate the instantaneous deformation from the rheological deformation. Therefore, it is necessary to comprehensively invert the parameters of the static constitutive model and the rheological model parameters. Using the measured displacement data, the static constitutive model and the rheological model parameters which are sensitive to the deformation of rockfill dam are parameters to be inverted. The particle swarm optimization algorithm based on particle migration and radial basis function neural network are used to construct the parameter inversion platform This method overcomes the shortcoming of particle swarm optimization, which is apt to fall into local optimum and premature convergence. The trained neural network is used to describe the mapping relationship between model parameters and displacements, and the computation time of parameter inversion is saved. The inversion results of Shuibuya high panel dam show that the calculated settlement values based on the inversion parameters are in good agreement with the measured values, and the deformation of the dam body is within a reasonable range and tends to be stable.